Ex‐vivo microbleed detection in community‐based older adults using confidence‐aware learning

• Published in: Alzheimer's & dementia Vol. 19; no. S12
• Main Authors: Nikseresht, Grant, Evia, Arnold M, Bennett, David A. A, Schneider, Julie A, Agam, Gady, Arfanakis, Konstantinos
• Format: Journal Article
• Language: English
• Published: 01.12.2023
• ISSN: 1552-5260; 1552-5279; 1552-5279
• DOI: 10.1002/alz.076785

Background Cerebral microbleed (CMB) detection on MRI scans of autopsied brains of community‐based older adults is crucial for MR‐pathology studies of cerebral small vessel disease (SVD). However, training an ex‐vivo CMB detection model is difficult due to the low prevalence of CMBs in the brains of community‐based older adults, and the fact that CMB mimics greatly outnumber CMBs on ex‐vivo MRI. We demonstrate that confidence‐aware deep learning, a technique for enforcing correct ranking of examples during neural network training, significantly improves ex‐vivo CMB detection performance and enhances the interpretability of model predictions. Method CMBs from 286 participants from two longitudinal cohort studies of aging were included in this work. T2*‐weighted gradient echo scans of autopsied brains with a voxel size of 1×1×1 mm were used. Confidence‐aware learning was used to regularize deep learning model predictions such that they reflect the degree of confidence the detection model has in a prediction. Model confidence was estimated during training by correctness, the number of correct prediction events across epochs divided by the total number of epochs. The objective loss function was modified to include a correctness ranking loss (CRL) term that encourages correspondence between correctness and predicted probabilities. Models are trained with and without the CRL loss term to evaluate the impact of confidence‐aware learning on model performance. An end‐to‐end CMB detection framework that combines data synthesis, candidate selection, false positive reduction, and full scan evaluation was used as the backbone for this work. Result CMB detection models trained using correctness ranking loss achieved a significantly higher ensemble average precision (0.2915) compared with models trained without correctness ranking loss (0.2379). The CRL trained model achieved a 25% higher sensitivity at 0.5 false positives per subject and 12% higher sensitivity at 3 false positives per subject when compared to models trained without CRL. Conclusion This work demonstrates that training with confidence‐aware learning can improve the performance and interpretability of ex‐vivo CMB detection algorithms in community‐based cohorts. The use of a confidence‐aware CMB detection algorithm for assisted CMB annotation in ex‐vivo MRI will aid in future MR‐pathology studies into the relationships between CMBs and neuropathology.